From mboxrd@z Thu Jan 1 00:00:00 1970 From: Wu Fengguang Subject: Re: [RFC][PATCH v3] readahead: introduce O_RANDOM for POSIX_FADV_RANDOM Date: Tue, 5 Jan 2010 10:16:53 +0800 Message-ID: <20100105021652.GA29428@localhost> References: <20091225000717.GA26949@yahoo-inc.com> <87aax18xms.fsf@basil.nowhere.org> <20091230051540.GA16308@localhost> <20091230052402.GB26364@localhost> <873a2s8hmp.fsf@basil.nowhere.org> <20100104045020.GA21021@localhost> <28c262361001032120v284e92b5ub1211f3d1fca6140@mail.gmail.com> <20100104121642.GA12266@localhost> <28c262361001041746j1270e2d2i79a932efca861dc5@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Andi Kleen , Andrew Morton , Quentin Barnes , "linux-kernel@vger.kernel.org" , "linux-fsdevel@vger.kernel.org" , Nick Piggin , Steven Whitehouse , David Howells , Al Viro , Jonathan Corbet , Christoph Hellwig To: Minchan Kim Return-path: Content-Disposition: inline In-Reply-To: <28c262361001041746j1270e2d2i79a932efca861dc5@mail.gmail.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Tue, Jan 05, 2010 at 09:46:09AM +0800, Minchan Kim wrote: > On Mon, Jan 4, 2010 at 9:16 PM, Wu Fengguang = wrote: > > Hi Minchan, > > > > On Mon, Jan 04, 2010 at 01:20:49PM +0800, Minchan Kim wrote: > >> > --- linux.orig/mm/readahead.c =C2=A0 2010-01-04 12:39:29.0000000= 00 +0800 > >> > +++ linux/mm/readahead.c =C2=A0 =C2=A0 =C2=A0 =C2=A02010-01-04 1= 2:39:30.000000000 +0800 > >> > @@ -501,6 +501,12 @@ void page_cache_sync_readahead(struct ad > >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0if (!ra->ra_pages) > >> > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0return; > >> > > >> > + =C2=A0 =C2=A0 =C2=A0 /* be dumb */ > >> > + =C2=A0 =C2=A0 =C2=A0 if (filp->f_flags & O_RANDOM) { > >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 force_page_ca= che_readahead(mapping, filp, offset, req_size); > >> > + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return; > >> > + =C2=A0 =C2=A0 =C2=A0 } > >> > + > >> > >> Let me have a dumb question. :) > >> > >> How about testing O_RANDOM in front of ra_pages testing? > >> > >> My intention is that although we turn off ra, it would be better t= o read > >> contiguous block all at once than readpage() callback doing I/O > >> one page at a time. > >> > >> Is it break some semantics or happen some problem in ondemand read= ahead? > > > > Yes it will have some problem with shrink_readahead_size_eio(), whi= ch > > want to disable readahead and use ->readpage() when ra_pages=3D=3D0= =2E > > > > Do you have specific use case in mind? The file systems that set > > ra_pages=3D0 seems to don't need readahead, too. >=20 > Never mind. It's just out of curiosity. :) >=20 > I thought although user disable readahead, we could enhance file I/O > with one readpages not multiple readpage if we know the user want to > read big contiguous blocks. Yes, not-break-large-read-into-pages would be good for HD/SSD drives when readahead is disabled. Currently, ->ra_pages is somehow overloaded in its =3D=3D0 case. As you said, it's in fact possible to disable readahead while still limiting read IO size to a non-zero ->ra_pages. > But I though it break current readahead off semantics. right? It can be done by applying the ->ra_pages limit to O_RANDOM. This also makes O_RANDOM safer to use: @@ -497,6 +497,13 @@ void page_cache_sync_readahead(struct ad struct file_ra_state *ra, struct file *f= ilp, pgoff_t offset, unsigned long req_size) { + /* be dumb */ + if (filp->f_flags & O_RANDOM) { + req_size =3D clamp_t(unsigned long, req_size, 1, ra->ra= _pages); + force_page_cache_readahead(mapping, filp, offset, req_s= ize); + return; + } + =20 /* no read-ahead */ if (!ra->ra_pages) return; To make real change, we need an interface for the user to disable whole-partition readahead by setting O_RANDOM instead of ra_pages=3D0. That would be a hard sell.. > Thanks for reply about my dumb question, Wu. :) You are welcome :) Thanks, =46engguang